梁 震

Researcher

Creating technology that connects people beyond language barriers

PhD Student at Kyoto University

OriginBeijing
CurrentKyoto

Messaging

Developer

Zenn

Social

About Me

About Me

Bridging Foreign Language Education and AI Technology

Education

PhD Student, Graduate School of Human and Environmental Studies, Kyoto University, Department of Cognitive Science in Language

PhD Student • Research

Experience

Teaching

Japanese Teacher at New Oriental (7 years, 5,000+ hours of teaching)

7+ Years Experience

Specialization

Foreign Language Education, Second Language Acquisition, Applied Linguistics

Large Language Model (LLM) Fine-tuning and Development

Machine Learning for Vocabulary Difficulty Prediction and Language Assessment

Educational Grammar and English for Academic Purposes

Academic Research

Research Activities

Innovating AI Technology in Language Education

Publications

A Study on Role Language in Manga Used in Daily Conversation: From the Perspective of Japanese Language Education

2025Peer-Reviewed Papers

Wang, F., Kanamaru, T., & Liang, Z.

Kotoba, 46, pp. 55–72

Analysis of high-frequency collocations based on English proficiency test textbook corpus

2024Conference Presentations

Nakano, T., Liang, Z., & Sasao, Y.

JASELE 49th Fukuoka Research Conference

Can general language models predict vocabulary difficulty based on Japanese learner data?

2024Conference Presentations

Liang, Z., & Sasao, Y.

Annual Meeting of the Association for Natural Language Processing 2024

Development and validation of an audio version of Japanese vocabulary size test for Chinese-Japanese bilinguals

2023Peer-Reviewed Papers

Peng, Y., Liang, Z., & Sasao, Y.

Journal of Japanese Language Education, 185, pp. 93–108

Motion and memory in VR: The influence of VR control method on memorization of foreign language orthography

2023Peer-Reviewed Papers

Vincent, N. H., Liang, Z., & Sasao, Y.

International Journal on Cybernetics & Informatics (IJCI), 12(1), pp. 151–164

Use of subtitles in video content for Japanese language learning: From the perspective of language selection

2022Peer-Reviewed Papers

Peng, Y., Liang, Z., & Sasao, Y.

Language and Culture Education Research, 20, pp. 335–356

Development and validation of an audio version of Japanese vocabulary size test for Chinese-Japanese bilinguals

2022Conference Presentations

Peng, Y., Liang, Z., & Sasao, Y.

Proceedings of the Autumn Conference of the Society for Teaching Japanese as a Foreign Language

Development and verification of an automatic choice generation program for Japanese vocabulary questions

2022Conference Presentations

Liang, Z., & Sasao, Y.

Proceedings of the Spring Conference of the Society for Teaching Japanese as a Foreign Language

技術的専門性

スキル・技術

フルスタック開発から機械学習まで

プログラミング言語

JavaScript
95%
TypeScript
90%
Python
85%
Swift
80%

フロントエンド

React
95%
Next.js
90%
React Native
85%

バックエンド

Node.js
90%
Django
85%

データベース

MongoDB
90%
MariaDB
85%

クラウド

AWS
85%
Google Cloud
80%
Vercel
90%

AI・機械学習

PyTorch
85%
TensorFlow
80%
資格・認定

専門資格・達成実績

2020年

JLPT N1(満点)

日本語能力試験 - 最高レベルで満点取得

2019年

ICTプロフィシエンシー検定

情報通信技術に関する資格

2022年

日本運転免許証

日本で発行された普通自動車免許

2018年

中国運転免許証

中国で発行された普通自動車免許

2018年

大学英語テスト4級

中国の大学生向け全国英語能力試験

2025年

CATTI国際版 翻訳

中日翻訳認定証 - 国際通用翻訳能力等級試験(CATTI国際版)

2025年

CATTI国際版 通訳

中日通訳認定証 - 国際通用通訳能力等級試験(CATTI国際版)

言語能力
🇯🇵

日本語

JLPT N1 満点・ネイティブレベル

習熟度100%
🇨🇳

中国語

母国語(北京出身)

習熟度100%
🇺🇸

英語

CET Level 4・研究論文執筆レベル

習熟度75%
Featured Projects

Project Portfolio

Practical Application Development

Tech Leadership

Spearheaded Agile development methodology implementation, CI/CD pipeline architecture, microservices design patterns, and DevOps culture cultivation. Comprehensive oversight from product management to technical architecture decisions.

Agile/ScrumCI/CDMicroservicesDevOpsProduct ManagementTechnical Architecture

Sokudoku Gorilla

Ongoing2025-Present

Interactive speed reading practice system for Japanese learners

Main Features

  • Level-based content from beginner to advanced
  • Real-time reading speed (WPM) and comprehension measurement
  • Interactive learning with quiz format
  • Progress visualization and tracking

Technology Stack

Next.js
TypeScript
Tailwind CSS
Vercel
PostgreSQL

日雨日語 (Nichiu Nichigo)

Completed2022-2023

Efficient Japanese vocabulary learning and listening practice mobile application

Main Features

  • Cross-platform support with React Native and Swift
  • Flexible data modeling with MongoDB
  • Serverless processing on AliCloud

Technology Stack

React Native
Swift
Node.js
MongoDB
AliCloud Function Compute

Vocabulary Profiler 'Matsunoha'

Active2023-2024

Web tool for text analysis that determines word difficulty and visualizes vocabulary levels

Main Features

  • Full-stack development with Next.js
  • MeCab morphological analysis on AWS Lambda
  • Robust code design with TypeScript

Technology Stack

Next.js
AWS Lambda
Node.js
Vercel Postgres
TypeScript

LandS English Learning App

Completed2021-2022

English learning application specialized in listening and speaking skills

Main Features

  • Robust infrastructure based on Google Cloud
  • Native experience implementation with Swift
  • High-speed response with Nginx

Technology Stack

Google Cloud Platform
Swift
Nginx
MariaDB

LLM Analysis & Fine-tuning Research

Ongoing2023-Present

Research project for analyzing and customizing LLMs specialized for foreign language education

Main Features

  • Large language model analysis with TensorFlow
  • LLM customization for foreign language education
  • Utilizing Transformers library

Technology Stack

Python
TensorFlow
Transformers
Axolotl

LLM Vocabulary Difficulty Estimation System

Ongoing2024

Development of a system that predicts vocabulary difficulty by training large language models with known difficulty data. Joint research project with National Institute for Japanese Language and Linguistics.

Main Features

  • Automatic vocabulary difficulty prediction using BERT-based models
  • Machine learning models utilizing diverse linguistic features
  • Industry-academia collaboration with National Institute for Japanese Language and Linguistics
  • Evaluation using Japanese learner corpus data

Technology Stack

Python
TensorFlow
Transformers
BERT
Natural Language Processing

Education-Specialized LLM Development Project

Ongoing2024

Research on developing education-specific large language models with novel architectures that overcome constraints of existing LLMs for foreign language education.

Main Features

  • Custom model architecture design for educational domains
  • Multilingual educational content generation
  • Learner-level adaptive text generation
  • Novel evaluation metrics development and validation

Technology Stack

PyTorch
Transformers
Hugging Face
Custom Architecture
Fine-tuning

Automatic Vocabulary Question Generation System

Ongoing2024

Advanced vocabulary question generation system for Japanese language education utilizing Word2Vec embeddings and Latent Dirichlet Allocation (LDA). Leverages distributional semantics and topic modeling for high-precision distractor generation.

Main Features

  • Semantic similarity extraction using Word2Vec distributed representations
  • Contextual distractor generation through LDA topic modeling
  • Vocabulary difficulty estimation via MeCab morphological analysis
  • Automated quality assessment using machine learning algorithms

Technology Stack

Python
Word2Vec
scikit-learn
Gensim
NLTK
MeCab
Loading articles...

Teaching Experience

Japanese Language Education

JLPT N1 Perfect Score

180/180

Highest achievement in Japanese proficiency

7
Years of Experience
5,000+
Teaching Hours
300+
Students Taught
95%
JLPT Pass Rate

Professional Teaching Background

New Oriental Education & Technology Group

As a senior Japanese language instructor at New Oriental, I have developed and delivered comprehensive Japanese language curricula for students at all levels, from beginners to advanced learners preparing for JLPT N1.

Key Achievements
  • Developed original teaching materials and methods
  • Maintained 95% student JLPT pass rate
  • Recognized as Top Instructor multiple times
Teaching Specialties
  • JLPT N1-N5 comprehensive preparation
  • Business Japanese for professionals
  • Conversational fluency development

Course Offerings

Basic Japanese Course

Foundation course for beginners starting their Japanese journey

  • Hiragana and Katakana mastery
  • Essential grammar patterns
  • Daily conversation skills
  • Cultural understanding

JLPT Preparation

Comprehensive test preparation for all JLPT levels

  • Targeted vocabulary building
  • Grammar pattern mastery
  • Reading comprehension strategies
  • Listening skill enhancement

Business Japanese

Professional Japanese for workplace communication

  • Business etiquette and keigo
  • Email and document writing
  • Meeting and presentation skills
  • Industry-specific terminology

Teaching Philosophy

Language learning is not just about memorizing words and grammar; it's about building bridges between cultures and opening doors to new opportunities. My goal is to make every student's journey both effective and enjoyable.

Ryo Shin (Liang Zhen) | AI Researcher・ML Engineer・Japanese Language Teacher